OpenAI's GPT-6 Astra Tops ErdosBench, But Prioritizes RSI Over Math
OpenAI's GPT-6 Astra scored 3.23 on ErdosBench, solving 106 math problems, including 43 fully, in its first major math-focused benchmark test.
Research
292 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
LATEST NEWS
-
-
AWS Introduces Agent Evaluation Metric for Multi-Turn Conversations
AWS announced the Agent Evaluation Metric (AEM), a turn-level tool to measure agent correctness in multi-turn conversations, with a focus on identifying root causes of failures.
-
OpenAI Uses Codex and ChatGPT to Search for New Antimicrobial Molecules
OpenAI's researchers are using Codex and ChatGPT to search for new antimicrobial molecules, reducing the initial discovery phase from years to hours.
-
OpenAI Faces Dispute Over AI-Generated Millennium Proof
OpenAI is accused of using researcher drafts and pressuring co-authors in a dispute over an AI-generated solution to a Millennium Problem, raising concerns about trust in AI labs.
-
Google Unveils AlphaGenome Atlas for Genome-Wide Variant Analysis
Google announced AlphaGenome Atlas on Tuesday, a system that evaluates all possible single-base changes in the human genome, analyzing 9 billion bases.
-
Pathway Introduces Brain-Inspired BDH Architecture on Amazon SageMaker HyperPod
Pathway’s brain-inspired BDH architecture improves reasoning efficiency by 150M parameters, achieving state-of-the-art results on ARC-AGI-1.
-
Google DeepMind Launches AlphaGenome Atlas for DNA Variant Analysis
Google DeepMind unveiled AlphaGenome Atlas, a tool that predicts the effects of 9 billion DNA variants, marking a major leap in biomedical research.
-
Danijar Hafner Develops Plan-Ahead Agents for Robotics
Danijar Hafner, 31, is developing plan-ahead agents for robotics, using model-based reinforcement learning to enable robots to navigate unfamiliar environments.
-
OpenAI Researcher Allegedly Pressured Mathematician to Drop Anthropic Co-Author from Math Breakthrough Paper
Mathematician Tristan Buckmaster claims OpenAI pressured him to remove Anthropic co-author Levent Alpöge from a Navier-Stokes paper, after the work leaked in early September 2026.
-
OpenAI Warns of AI Self-Improvement Pace as Research Interns Take Over
OpenAI reports its AI agents now handle three times the work of human researchers, with a 124-fold increase in token output since December 2025.
-
Insilico Medicine's AI Drug Shows Promise in Reversing Biological Age
Insilico Medicine's AI-designed drug rentosertib may reverse biological aging markers in early trial, with six aging clocks showing treated patients appeared up to six years younger than placebo group.
-
Hugging Face Launches Vaani Noise Dataset for Real-World Speech AI
Hugging Face released the Vaani Noise Event Timestamp Dataset, capturing 106,892 real-world noise events across 58 Indian languages on September 7, 2026.